CERVIXNET: An Efficient Approach for the Detection and Classifications of the Cervigram Images Using Modified Deep...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Cervical cancer detection relies on accurate image analysis.
• Early diagnosis significantly improves patient outcomes and survival rates.
• Conventional methods may lack the precision for early-stage detection.

Purpose of the Study:

• To develop a novel deep learning methodology for early cervical cancer detection.
• To segment cancerous regions within abnormal cervigram images.
• To enhance the accuracy of cervical cancer detection through a modified deep learning architecture.

Main Methods:

• A novel deep learning architecture, CervixNet, was proposed.
• The methodology involved training and testing phases using healthy and cancerous cervigram images.
• Cancerous regions were segmented using a specialized algorithm on abnormal images.

Main Results:

• The CervixNet architecture achieved high performance on the IMODT database: 98.69% Cancer Pixel Sensitivity (CPS), 98.76% Cancer Pixel Specificity (CPSP), and 99.27% Cancer Pixel Accuracy (CPA).
• On the Guanacaste database, results were: 99.22% CPS, 99.03% CPSP, and 99.01% CPA.
• Performance was evaluated against ground truth images using key metrics.

Conclusions:

• The proposed CervixNet methodology demonstrates significant novelty and effectiveness in cervical cancer detection and segmentation.
• Experimental results show superior performance compared to existing state-of-the-art methods.
• This approach holds promise for improving early detection rates and patient survival in cervical cancer.